a16z Partner: Code Is Infinitely Devaluing—Why Network Effect Is the Ultimate Moat of the AI Era?

Author: Eddy Lazzarin (CTO of a16z Crypto)

Compiled by: TechFlow

TechFlow Summary: The rise of open-source software once completely disrupted development paradigms with "composable code," allowing developers to rapidly integrate atop mature modules without reinventing the wheel. However, as large language models drive the marginal cost of writing software toward zero, "Vibe Coding" has drastically lowered the barrier to replicating application-layer interfaces, quickly dissolving the scarcity of code itself and eroding traditional software moats. Eddy Lazzarin, CTO of a16z Crypto, points out that the core of current productivity restructuring is shifting from reusable code to "reusable networks."

When software supply becomes abundant, what is truly scarce are foundational accumulated assets: deep liquidity, immutable reputation, interoperable identity graphs, and settlement trust verified by network-wide consensus. Blockchain serves as the canonical shared state layer of the internet, forming an irreplaceable structural moat for decentralized ledgers. While interfaces and code can be cheaply replicated, reusable networks that carry historical coordination and consensus value constitute the unshakeable ultimate moat of the agent era.

Open source has fundamentally reshaped the software industry by enabling composability in code. Developers no longer need to build everything from scratch; instead, they can directly assemble complete applications from existing codebases, frameworks, and protocols.

Today, as artificial intelligence drives the cost of producing software to zero, the core of composability is shifting: it is no longer solely about reusable code, but about reusable networks.

Software itself is no longer the moat; the network is.

Whether for the crypto space or more broadly for software developers and startups across industries, composability has always been one of the most compelling core concepts.

Early decentralized finance (DeFi) protocols (such as Uniswap) allowed new applications to be built directly on top of existing liquidity without having to rebuild liquidity pools from scratch, vividly demonstrating the immense power and vast potential of composability.

However, this trait was never truly the defining narrative of the crypto industry until now. The fundamental reason is that the trial-and-error and experimentation costs for innovation were too high.

Building on-chain applications requires highly specialized knowledge in smart contracts, security protocols, code auditing, and blockchain infrastructure. Exploring this vast design space consumes massive amounts of time and capital, which has always kept the pool of builders who could participate relatively small.

The arrival of AI has completely upended this dynamic. Today, anyone with an idea, including non-technical builders without a programming background, can generate a runnable prototype in hours, integrate existing protocols, and achieve product iterations that previously took months, all within a few hours.

Revisiting Crypto’s Most Ambitious Visions

• NFTs (Non-Fungible Tokens): Created digital-native assets, including collectibles and art, event tickets, real-world assets (RWAs), community identity credentials, and intellectual property (IP), all of which can be seamlessly recognized by applications and extended upon.

• DAOs (Decentralized Autonomous Organizations): Provided an unprecedented testbed for internet-native organizational collaboration and novel on-chain governance.

• Decentralized social networks: Unlocked fully portable, user-owned digital identities and social graphs.

• Onchain games: Transformed virtual worlds into truly open underlying infrastructure and thriving internet economies.

Only when massive numbers of real users and numerous AI agents choose to continuously build upon these networks will these ambitious visions truly unleash their immense power.

AI can essentially recreate and generate virtually any application. It can clone front-end interfaces, rewrite API endpoints, and produce entirely new software products in minutes. AI not only makes software development extremely cheap, but it also fundamentally upends the economic definition of scarcity.

While software is becoming abundantly plentiful, networks themselves certainly are not.

Consider this scenario: if building crypto wallet software becomes as easy as clicking a button, traders could instantly generate personalized wallets tailored to themselves; and as AI agents directly execute on-chain transactions on behalf of users, traditional wallet software interfaces might disappear entirely.

In such a future world, the importance of front-end interfaces will pale in comparison to the underlying networks they rely on. The truly scarce asset is no longer the "best wallet client," but rather access to the deepest liquidity pools, the strongest historical reputation backing, the richest identity relationship graph, and the most trusted settlement network across the network.

AI can easily generate an NFT, but it absolutely cannot conjure out of thin air the provenance, community consensus cohesion, and extensibility to other external networks behind that specific NFT.

AI cannot replicate your on-chain digital identity, credit reputation, or social graph out of thin air, nor can it generate the massive liquidity accumulated within protocols. It certainly cannot replicate the historical accumulation, multi-party coordination, and consensus trust that a network has cultivated over the years.

You can never create a USDC out of thin air relying solely on "Vibe Coding."

But the Internet Needs Shared State

Perhaps this is precisely the ultimate mission blockchain was destined to fulfill from its inception: to become the most authoritative and canonical shared state layer across the entire internet.

As upper-layer applications gradually become commoditized, the value capture point of the entire tech ecosystem will inevitably shift comprehensively toward the networks these applications are attached to and composed with. In a future world filled with AI agents, this unified global shared state is more indispensable than ever before.

Front-end interfaces can be instantaneously copied and cloned, and software code can be repeatedly and quickly refactored by AI. However, the core underlying state—including asset ownership, identity verification, liquidity depth, reputation systems, and network collaboration relationships—still possesses absolute scarcity that cannot be diluted.

For today’s developers and entrepreneurs, these "reusable networks" with strong network effects represent the most grand and historic opportunity of the AI era.